Fix record offsets when parsing a list of PAC records

[kwy404]

PacRecord.list_from_bytes moved to the next record by adding len(bytes(record)), which is the length of the re-serialized record, not the number of bytes that were parsed. CodecSpecificCapabilities.__bytes__ always writes all five LTVs, but Supported_Audio_Channel_Counts and Supported_Max_Codec_Frames_Per_SDU are optional and default to 1 when absent (BAP 4.3.1). A record without them is 6 bytes shorter than its re-serialized form, so every record after it in a Sink or Source PAC value (PACS Table 3.2 and 3.4) was parsed from the wrong offset. This affects PublishedAudioCapabilitiesServiceProxy.sink_pac and source_pac, used for example by apps/device_info.py.

This adds PacRecord.parse_from_bytes(data, offset), following the CodingFormat.parse_from_bytes pattern, which returns the offset after the record, and uses it in list_from_bytes. from_bytes keeps its behavior.

Added test_pac_record_list in tests/profiles/bap_test.py, which parses two LC3 records that only carry the mandatory capabilities; it fails before the change (the second record is decoded starting 6 bytes too late and raises UnboundLocalError) and passes after. Ran the tests/profiles suites, plus black, ruff, pylint and mypy on the touched files.

[zxzxwu]

LGTM, thanks!